Research on Applying Multimodal Large Language Models to Assist C++ Instruction

Abstract

This research addresses prevalent issues in traditional C++ programming instruction, such as monotonous learning methods, the disconnect between theory and practice, and the lack of a robust assessment framework. It proposes an innovative AI-assisted teaching system that integrates multimodal large language models (MLLMs) with digital human technology. The system establishes an intelligent multimodal interaction framework, a dynamic knowledge graph, an adaptive learning engine, a real-time code interaction module, and a practical task and progress tracking mechanism. This integrated approach enables comprehensive coverage of programming concepts, personalized instruction, and real-time code diagnostics, ultimately enhancing students' practical programming skills. A survey-based methodology was employed to evaluate the system's projected effectiveness. The results indicate that the system holds significant potential for improving learning efficiency, enriching the learning experience, and facilitating deeper knowledge acquisition. Consequently, it presents a promising solution for innovating C++ programming pedagogy.